Extraction of Character String Regions from Scenery Images Based on Contours and Thickness of Characters
Tomohiro Nishino, Kimitoshi Yamazaki, Kei Okada · 2011
This paper describes a method to find strings of characters from natural scene images. It finds character candidates by taking particular note of closed contours in an image, and detects character strings by evaluating their size and line consistency. In the case of low resolution images, some disadvantageous condition occurs, for instance, only a part of actual contour may be able to be found or two closed edges are merged between two characters. To cope with such difficulties, character candidates are first selected, and then the uniformity of thickness of them is evaluated. Character strings can be found through this process. After that, additional character candidates are searched again around the already detected candidates. Moreover, the detection results of character strings are used to find rotation parameters to transform image for reading the character string with propriety. 1